Open to opportunities

Sarthak Shandilya

@sarthakshandilya

Message

Data Scientist and ML Engineer delivering predictive, time-series and generative ML solutions.

Zimbabwe

Message

What I'm looking for

I’m seeking full-time Data Scientist or ML Engineer roles where I can deliver data-driven insights and scalable ML solutions—building models end-to-end, validating with strong metrics, and shipping deployment-ready systems.

I’m a Data Scientist and ML Engineer with end-to-end project experience spanning predictive modelling (AUC ~0.97), time series forecasting (~60% error reduction), and deep generative modelling (Conditional DCGAN, WGAN-GP).

I build models that don’t just perform in notebooks—I focus on stability, reproducibility, and deployment. For example, MorphGen trained a Conditional DCGAN on CelebA (~200K images), used WGAN-GP with gradient penalty and spectral normalisation, and achieved stable convergence without mode collapse.

On the forecasting side, I delivered DemandIQ with ARIMA, SARIMA, Prophet, and XGBoost over 1M+ retail records, reducing forecast error ~60% vs. a naive baseline. I also containerised pipelines with Docker, tracked experiments with MLflow, and exposed insights via Streamlit for non-technical stakeholders.

For GTM analytics, PortIQ combined EDA and SQL on 1,200 AIS vessel records with 27 domain-specific GTM features, reaching AUC ~0.97 (5-fold cross-validation). I’m also open-sourcing my workflows (e.g., MorphGen) and using disciplined experimentation practices through MLflow to support scalable, data-driven decision-making.

Experience

Work history, roles, and key accomplishments

Current

Conditional Generative Face Synthesis

Current

MorphGen

Apr 2026 - Present (2 months)

Trained a Conditional DCGAN on the CelebA dataset (~200K images) using WGAN-GP loss with gradient penalty and spectral normalization. Delivered attribute-conditioned face synthesis (gender, age, expression) with stable convergence over 50 epochs and deployed interactive real-time inference via Streamlit.

PyTorch Conditional DCGAN WGAN GP Gradient Penalty Spectral Normalization CelebA Label Conditioning Streamlit

Current

GTM Opportunity Scoring Platform

Current

PortIQ

Mar 2026 - Present (3 months)

Performed EDA and SQL-based querying on 1,200 AIS vessel records and engineered 27 domain-specific GTM features. Built a Random Forest classifier achieving AUC ~0.97 with 5-fold cross-validation across 180 companies, and automated tiered prospect ranking with MLflow-tracked experiments.

Random Forest EDA SQL Feature Engineering AUC Optimization Cross Validation MLFlow Streamlit

Current

Retail Demand Forecasting & Risk

Current

DemandIQ

Jan 2026 - Present (5 months)

Forecasted demand across 1M+ retail records using ARIMA, SARIMA, Prophet, and XGBoost, reducing forecast error ~60% vs. a naive baseline. Implemented Dockerized PySpark pipelines with MLflow tracking and delivered anomaly detection, scenario simulation, and confidence intervals via Streamlit.